Sign in Subscribe
Machine learning

Machine Learning Breakthrough in Molecular Dynamics Simulations

Strategic Insights

— 1 min read

Researchers from the Ames Laboratory, led by Weiyi Xia, have made a significant breakthrough in machine learning, developing an accurate and transferable artificial neural network machine learning (ANN-ML) interatomic potential for molecular dynamics simulations of the La-Si-P system. This study demonstrates the crucial role of training data in ML potential development and accurately describes the energy vs volume curves for various crystalline and liquid structures in the La-Si-P system.

Key Takeaways:

  • The developed ANN-ML potential accurately describes the energy vs volume curves for all known elemental, binary, and ternary crystalline structures in the La-Si-P system.
  • The ANN-ML model systematically underestimates the melting temperatures of several crystalline phases in the La-Si-P system, but the overall trend agrees with experimental observations.
  • The research demonstrates the importance of training data in ML potential development and the potential of ANN-ML in accurately describing complex materials' behavior.
  • The study was funded by the U.S. Department of Energy and the National Natural Science Foundation of China.
  • The research has been peer-reviewed and published in The Journal of Chemical Physics, a publication of the American Institute of Physics.

Statistics:

  • The developed ANN-ML potential was trained on a dataset of 100,000 structures, comprising elemental, binary, and ternary crystalline structures in the La-Si-P system.
  • The ANN-ML model was applied to study the nucleation and growth of LaP as a function of different relative concentrations of Si and P in the La-Si-P liquid, with results consistent with experimental observations.
  • The study predicted the melting temperatures of several crystalline phases in the La-Si-P system, with a systematic underestimation of 10-20% compared to experimental values.

Sources:

  • Xia, W., Tang, L., Viswanathan, G., Soto, E., Kovnir, K., & Wang, C.-Z. (2025). Developing a neural network machine learning interatomic potential for molecular dynamics simulations of La-Si-P systems. The Journal of Chemical Physics, 163(8).
  • NewsRx. New Machine Learning Data Have Been Reported by Researchers at U.S. Department of Energy (Developing a neural network machine learning interatomic potential for molecular dynamics simulations of La-Si-P systems). Journal of Engineering. September 15, 2025; p 1553.

Read more